Real coding sessions · never self-reported
Compared two agent workflow tools through recordings and reviewer tests
First build tracked for this project
Compared Panels with bb using screen recordings and three reviewer runs, which leaned toward Panels with conditions.
Explored thr_2864esqmed
2.3M tokens vs typical 197.4K
1h 14m session reading and analyzing the codebase. 1.0M tokens across 22 prompts using claude-opus-5-5.
Explored thr_2864esqmed
2.1M tokens vs typical 197.4K
1h 14m session reading and analyzing the codebase. 1.0M tokens across 22 prompts using claude-opus-5-5.
Explored thr_2864esqmed
24 prompts
1h 14m session reading and analyzing the codebase. 1.0M tokens across 22 prompts using claude-opus-5-5.
Explored thr_2864esqmed
5737 sessions tracked this month across all builders
10m session reading and analyzing the codebase. 527K tokens across 12 prompts using claude-opus-5-5.
Explored thr_2864esqmed
462 sessions tracked this week across all builders
10m session reading and analyzing the codebase. 527K tokens across 12 prompts using claude-opus-5-5.
Explored thr_2864esqmed
Previous best: 2,577 lines
10m session reading and analyzing the codebase. 527K tokens across 12 prompts using claude-opus-5-5.
Explored thr_2864esqmed
460 sessions tracked this week across all builders
10m session reading and analyzing the codebase. 527K tokens across 12 prompts using claude-opus-5-5.
Explored thr_2864esqmed
820 sessions tracked this week across all builders
10m session reading and analyzing the codebase. 527K tokens across 12 prompts using claude-opus-5-5.
Compared two agent-supervision tools and mapped where each fits
First build tracked for this project
Reviewed two agent tools' docs and a live instance, then picked a few features that would cut supervision effort.
Debugged thr_jncb3auvzv
5730 sessions tracked this month across all builders
22m session running commands and testing. 296K tokens across 16 prompts using claude-opus-5-5.
Debugged thr_jncb3auvzv
455 sessions tracked this week across all builders
22m session running commands and testing. 296K tokens across 16 prompts using claude-opus-5-5.